Closed Wauplin closed 2 weeks ago
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update.
Infra/hub team is looking into it (see slack thread). It should now be fixed server-side. @amyeroberts can you let me know if the problem persists in transformers
's CI? In the meantime, I'll convert this PR to draft and drop it if the fix is confirmed.
@Wauplin It seems that we can close this - I haven't seen any read time outs in the past 1.5 days since the fix was handled by the infra team. Thanks for responding to this so quickly!
Great news! Thanks for following-up on this. I'm closing it but feel free to reach out if it ever happens again :)
Related to internal slack thread (private link) cc @amyeroberts
Some tests started to fail in
transformers
. Might be related to an infra update or simply to a higher load in the CI. Failures are usually due to aReadTimeout
when downloading a file. There is already a retry mechanism when streaming the data from the server but not when making the first call. This PR fixes this by adding the same retry mechanism on the first call as well.Expected result: no more random failures in the CI.
(note: code is quite duplicated but I did not found a way of factorizing it nicely. So let's say it's fine and better for readability)